Because you don't win fantasy contests by racking up some arbitrary player rating points; you win by scoring the most daily fantasy points. So it doesn't really make sense to say
Quote:
One is that they’re quite fragile, requiring an unapproachable level of certainty to trust.
This makes it seem like there's some
other thing that is not fragile and
does have an approachable level of certainty to trust (whatever that means). But if such a thing existed, it would have to be highly-correlated to actual fantasy scores, or at least more correlated to the actual than whatever model(s) it is outperforming. If that's the case, then you could just add it to a model that predicts actual scores and see dramatic improvement in the estimates and prediction intervals.¹ If these ratings are any good at beating dfs, then they are basically just transformations of accurate point projections.
Quote:
The second is that our methodology is based upon providing you the tools you need to create (and backtest) your own models and generate your own player ratings.
I don't understand the logic here. Why can't you backtest models of predicted fantasy points against actual ones? He says this is the second reason why they don't offer point projections, but there's absolutely no methodological issue in backtesting the actual fantasy point values.
¹Assuming the model is correctly specified.